Dislocations and micropipes
associated with
hexagonal voids
Voids moving through the SiC boule can be expected to affect the distribution of defects. In particular, it is easy to document effect of void motion on threading edge and screw dislocations. Majority of threading dislocations in SiC boules are of the edge character. Since these are difficult to image using x-ray topography, the best way to monitor their distribution is by KOH etching. Fig. 1 shows two optical microphotographs of two wafers cut from the same SiC boule.
(a) (b)
Fig. 1 Nomarski contrast microphotographs of
two KOH etched SiC wafers cut from the same boule. Dark dots and lines correspond to intersections of threading
dislocations with the wafer surface. (a) Surface above the void seen as a
bright elongated hexagon (b) Surface of the wafer below the void.
Both microphotographs have been taken from the same part of the crystal as clearly visible from the appearance of two agglomerations of defects in upper left corner of images. The image on the left corresponds to the surface above the void which can be seen as a bright elongated hexagonal outline. Dislocation etch pits and dislocation arrays show no correlation with the location of the void. Image on the right was taken from the wafer just below the void. The void path is clearly outlined with rows of dislocation etch pits. The lower right corner of the void path has two large hexagonal pits often interpreted as due to the micropipes. Also, the interior of the re-crystallized portion has lower dislocation density then the surroundings.
Two etch patterns shown in Fig. 2 provide additional insight into the process of void motion and defect re-distribution. Both images were obtained on one KOH etched on-orientation wafer. In Fig. 2(a), the microscope was focused on the void contained within the wafer. The etch pits on the bottom surface of the wafer are visible as dark diffuse lines shifted toward upper right corner. This shift indicates that the void moved not only along the c-axis during growth but also laterally. Lateral motion was directed away from the center of the boule toward the crystal periphery. This is in agreement with the temperature distribution within the boule. In PVT growth, the heat is extracted through the center of the seed and, as a consequence, the temperature is increasing toward growth surface and the outer periphery of the boule. Void, therefore, was moving up the temperature gradient. The nucleation on the bottom surface of the void should occur at the coldest opposite direction of lateral motion. The trench surrounding lower edges of the void corresponds then to the last solidified material.
(a) (b)
Fig. 2 Nomarski contrast optical microphotographs of (a) void within the SiC wafer located close to the crystal periphery and (b) etched bottom surface of the same wafer.
Fig. 2(b) shows the etched bottom surface of the same wafer. The etch pits line up only underneath the trench with large hexagonal pits located under the deepest parts of the trench. This is a characteristic behavior observed on most voids located close to the outer periphery of the boule – the dislocations are associated with the last to solidify bottom of the void that is also a leading edge in lateral motion.
(a) (b)
Fig. 3 White beam synchrotron radiation topographs of corresponding areas in two SiC wafers (back-reflection, (g = 00024, l = 1.24 Å) (a) the area above the void and (b) area below the void.
The tubular voids extending below the corners of the voids can be of the same origin as micropipes i.e. they can be stabilized by the presence of a super-dislocation with the large Burgers vector. The support of this hypothesis was obtained by White Beam Synchrotron Radiation Topography shown in Fig. 3. The two topographs can be easily aligned due to the presence of two large white circles corresponding to micropipes penetrating both wafers. The small white dots distributed more or less uniformly over the area are the images of elementary screw dislocations with Burgers vector 1c[0001]. The threading edge dislocations responsible for most of the etch pits in images above do not produce discernible contrast in this reflection. The white hexagon in the topograph on the left marks the position and shape of the void that was observed by optical microscopy in the wafer on left. It is easily seen that there are about 15 screw dislocations located above the void and most likely intersecting its top surface. There was no contrast associated with the void edges in this topograph.
The topograph in Fig. 3(b) obtained on a separate wafer shows two significant differences as compared to Fig. 3(a). One is the visible outline of the void path. The boundary appears dark on lower right of the void path and light on upper left. This indicates the slight tilting of the basal plane underneath the void. Second difference is the absence of elementary screw dislocations within the boundaries of the void path. Instead, there are several large diameter circular white contrast features located at the boundaries. Since this type of contrast is well documented as due to super-screw dislocations, the void passage served as a conduit for redistribution screw dislocations and formation of micropipes.
Fig. 4 is a schematic representation of the void and micropipe formation process. The PVT-grown wafers used as seeds typically contain 104-105 cm-2 threading edge dislocations and 103-104 cm-2 elementary screw dislocations. As void forms at the interface it intersects dislocations, which now terminate at the top of the void still open to the back of the crystal. The void begins to close by nucleation occurring at the sidewall. Initially this lateral growth should be almost dislocation free as only the basal plane dislocations could replicate into the overhang and other dislocation nucleation mechanisms such as stresses, inclusions, or high super-saturation are absent.
Fig. 4 Schematic diagram showing void formation and movement
It is well known that dislocation lines cannot end inside a crystal; they must terminate at another dislocation or a free surface. Although the interior of a void may be considered a free surface, an enclosed cavity in a crystal must have a net Burger’s vector of zero. If the cavity surface is divided in two (in our case the top and bottom void facets), then the net Burger’s vector of each half must be equal. Consequently, if the net Burger’s vector of dislocations terminating at the top of the void is not zero, then dislocations must be created in the re-crystallized material beneath the void. This is likely to occur in the last to solidify as observed in Fig. 2. In addition, the cantilevered growth front of the closing void bottom is thin and could flex up or down under the influence of small forces, which in turn would cause the formation of additional dislocations not already present in the seed. Essentially identical effect was observed in the lateral epitaxial overgrowth of GaN where cantilevered lateral growth fronts contact to produce rows of dislocations.
Based on the above argument one can expect the array of dislocations forming in the last to fill part of the void bottom. The strain associated with these dislocations increases the bond energy near the dislocation core, which in turn should lower the probability of adatom attachment. The growth rate near a dislocation is therefore expected to be slower. This leads to the formation of a trench opposite to the nucleation site. As the trench gets deeper, the temperature at its bottom will be lower than the rest of the void. Local increase of super-saturation will eventually cause the trench to close.
The depressions that occur at the corner of the void, formed by the last closing regions of the growth front in the trench, provide the location for screw dislocations to combine. If screw dislocations are in the region of the depression when it begins to close, then it is energetically more stable for the dislocation core to remain open because the surface energy of the open core is less than that of the strain energy of the closed core. The image forces associated with the large growth steps can push additional screw dislocations lined up in the trench along the edge of the void closer to the depressions at the void corners. Same-sign dislocations can combine in the hole, thus forming a micropipe at the bottom of a trench extending below the void. Fig. 5 is an optical micrograph in reflection of a void cross-section where a micropipe extends below the base of a trench formed by passage of a void, providing support for the model above.
Hexagonal voids approximately 300 mm on a side moving in a crystal with an elementary screw dislocation density of 1x104 cm-2 will intersect approximately 10 elementary screw dislocations. The results presented above show that elementary screw dislocations result in micropipes that extend beneath the void. If we assume that there are 3 micropipes created per void, and then consider the density of hexagonal defects in the wafers (1-5 cm-2), it is possible to have micropipe densities between 3 and 15 cm-2 due to motion of hexagonal voids. In addition, if one assumes uniform distribution of hexagonal voids throughout the boule length of one inch (50 times greater than the wafer thickness), then it is possible that wafers near the seed have micropipe densities between 150 and 750 cm-2. Since the micropipe density in currently available wafers is 10-100 cm-2, voids could be the dominant formation mechanism for micropipes. Elimination of hexagonal voids by use of a backside evaporation barrier and proper seed mounting techniques was shown to reduce micropipe densities in PVT grown SiC (Sanchez et al. J. Elec. Mat. 29, 347 (2000)).
Return to:
[ CMU SiC Page] [ MSkowronski Home Page ]
